Learning Forum

[fares abbara]

I am running a basic system coupling to simulate tube flow simulation using fluid solid interface. when reducing the time step size i encounter instabilities that result in unrealistic deformations. when replaying deformations in the tube in transient structural the tube looks like it is exploding. with larger time steps this issue does not exist and the simulation runs smoothly. My question is, why does this instability exist with reduced time stepping and how can i overcome that issue ?

[fares abbara]

I am really confused as to why my question is not at least being viewed. This is the 3rd time i post a question regarding this issue and yet no ansys employee seems to have at least a remark over this issue. kindly let me know how can i rephrase my question so that it would be more clear for anyone to understand. 

Thank you.

[Rob]

This thread?  https://forum.ansys.com/forums/topic/time-stepping-issue-when-simulating-flow-in-an-elastic-tube/ 

[fares abbara]

Hello Rob,

Thank you for your reply.

Yes, in this thread that you mentioned i showed how the the total displacment of the tube was unreasonalble, almost as if the tube is exploding. This issue happens only when i reduce the time step size to 0.001. I need to reduce the time step size to this amount since i need to capture a cyclic loading i am planning to introduce on the tube. I am not sure what the nature of this instability is. After doing my own research this issue in FSI is not uncommon, and it has been suggested, to my understanding, that combining a coarse mesh with small time step size can result in such instabilities. In the system coupling transcripts (when running the simulation) it is visible that the forces are not converginging at the interface.

[fares abbara]

For better understanding of my model, here is a summarized problem statement.

Model Specifications are as follows:

 

• Linear elastic tube with young’s modulus 0.108 MPa and 0.49 Poisson ratio.
• Tube thickness 1mm and inner diameter 8.5mm.
• Non-Newtonian fluid model.
• Outlet boundary condition: 1300 Pa.
• Inlet boundary condition: time dependent sinusoidal velocity (amplitude 0.1 m/s, frequency 2Hz).

The model runs without errors with time step of 0.01s. Model becomes highly unstable with unreasonable deformations in the tube when the simulation runs at 0.001s time step size. (other time steps lower than 0.01 also cause failure).

 

I have tried to:

• Check the contact between the solid and fluid domain for any gaps that could be causing instability.
• Experiment with constant and time dependent linear inlet velocity profile.
• Running the simulation in steps (run fluent first and then export data as initial conditions for coupled analysis).
• Experimented with numerical stabilization controls in the structural domain.

[Rob]

With a flow pulse at 0.1s period I'd expect convergence issues at 0.01s time step, so it's more likely that you're picking up something real at 0.001s step size. If you run fixed mesh how does the pressure field look with time? NonNewtonian fluids do interesting things to the flow, so you may find there's an unresolved flow feature somewhere. 

You may also want to review the Fluent intrinsic FSI as the solid material properties look "simple" so you'll save time on the system coupling. There's a tutorial/video in Help which should explain most of it. 

As an aside for the first two posts. 26th August was a Saturday, so you're not going to get an answer from Ansys staff: we generally don't work weekends. Bumping a thread after 5-6 hours won't speed up a response, and may delay it as it messes with one of the (Ansys) check scripts and is more likely to annoy (nonAnsys) people who may help rather than encourage the community to step in. 

[fares abbara]

Thank you for you reply Rob.

Regarding my previous post, I apologise for my lack of knowledge in the way posts are managed and not being more patient, as I have struggled with this issue for quiet a while now, which has drastically crippled my research, and never really got to a clear solution or explanation. I am learning along the way. 

Going back to my model, if the flow pulse is at 2Hz wouldn't that make the period 0.5s?

 With a fixed mesh the pressure field appears to be banded along the length of the tube. The bands are very close is their pressure readings. This is something that was confusing to me as i ran the simulation for only 0.1 seconds and was expecting to be so pressure variation along the length of the tube. 

Finally, would you kindly direct me to the video you are referring to? do i get it using the help tab on workbench?

Thank you again and i extend my most sincere appologies. 

[Rob]

Oops, I read too many posts and picked up the 0.1 and Hz whilst missing the bit in the middle. At 2Hz (0.5s cycle) you will want 20+ time steps per cycle as you need to resolve the change in flow/pulse/whatever: if the speed/cell size needs a smaller step size that's your limit. For high quality results 100 steps isn't unreasonable, and you will also need some cycles to flush the initial solution out of the domain if you're after a "equilibrium" cycle. 

If you're seeing banded pressure on the fixed mesh do some further investigation to see why. It may be the fluid viscosity, but could be you're seeing something else. If you post images someone may be able to give you some pointers. 

The videos are in the Fluent Help system, so, yes, click on Help on the main screen. Help on the panels will take you to specific pages, which can also get you to the top level. 

No worries. The trick with support is to have a proper think and then ask a question just before becoming frustrated. Questions aren't "daft" if you've had a think and looked in Help/searched on here; and doing that also avoids too frequent asking without wasting your time being stuck. Staff only know what we know because we've been doing this a while, and have a collective experience that's measureable in decades/centuries.